Hybrid ARMA-GARCH Neural Networks for Financial Time Series: Application to Russian ESG Stock Returns (2013-2023)

Réseaux neuronaux hybrides ARMA-GARCH pour les séries temporelles financières: application aux rendements des actions ESG russes (2013–2023)

Sarah Goldman
Additional contact information
Sarah Goldman: Lux-SIR, CRIISEA - Centre de Recherche sur les Institutions, l'Industrie et les Systèmes Économiques d'Amiens - UR UPJV 3908 - UPJV - Université de Picardie Jules Verne

Working Papers from HAL

Abstract: The objective of this paper is twofold. First, it describes a family of nonlinear Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models for forecasting financial asset volatility. The second objective is to discuss the debates on the merits of using mixed GARCH neural networks in such research areas. For this purpose, we use different specifications, such as the ARMA-GARCH procedure and the ARMA-GARCH neural network (NN) approaches. We chose to model the daily Russian Sustainable Vector Index (RSVI) in RUB, covering the period 01/08/2013-08/22/2023, for at least two reasons. Given the environmental context and, more specifically, the urgency of climate change, various countries have signed the 2015 Paris Agreement. Russia has been applying it since November 2019. According to the STATISTICA website, "Russia's territorial carbon dioxide emissions per capita grew by 20 percent between 2000 and 2021"; however, in a very recent period, we have observed a decrease in CO2 emissions (for the period 2021-2022 -5.8%). Without sustainable support, it is impossible to reach the program's environmental transition calendar (i.e., net zero carbon for 2060). Therefore, all financial entities, especially financial markets, should participate in cleaning up the productive sphere. Secondly, very few econometric works are dedicated to the RSVI variable, and this analysis has tried to fill this deficiency by feeding the current reflection of the technical/computational and financial research. Therefore, we have run different specifications, including or not including neural network algorithms. We have used a machine learning approach and, based on two well-known performance tools (Root Mean Squared Error and Mean Absolute Error), we have concluded that the forecasted volatility of daily RSVI returns is improved using an ARMA-GARCH-NN specification. This conclusion is in line with the recent related literature.

Keywords: ARMA-GARCH process; Neural Network algorithms; Stock Market; Sustainability; Volatility forecasts (search for similar items in EconPapers)
Date: 2024-09-30
Note: View the original document on HAL open archive server: https://hal.science/hal-05241097v1
References: Add references at CitEc
Citations:

Downloads: (external link)
https://hal.science/hal-05241097v1/document (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:hal:wpaper:hal-05241097

Access Statistics for this paper

More papers in Working Papers from HAL
Bibliographic data for series maintained by CCSD ().